Performance counters are often used in computing systems in order to provide insight into system behavior. For example, system performance engineers may monitor and analyze data collected by performance counters in order to tune system performance. In computing systems that span networks, it is difficult to correlate performance counters from different devices. In particular, although the computing devices of the system may be periodically requested to sample their corresponding performance counters, those samples generally represent behavior that occurred during different time periods. As such, performance counter analysis has been traditionally limited to the analysis of counters that are tracked over many samples, over a relatively large period (e.g., 100 seconds), and/or averaged. Although doing so may reduce the errors associated with the skewed start and sampling times of the various computing devices, fleeting events are often entirely lost or suppressed in the data and the corresponding analysis.